Research map / 3D vision / Face and shape
GeLaTO: Generative Latent Textured Objects
Ricardo Martin-Brualla, Rohit Pandey, Sofien Bouaziz, Matthew Brown, Dan B Goldman
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use GLO latent optimization instead of VAE for instance latents (ablation) | GeLaTO · Eyeglasses view interpolation | PSNR: 39.70 -> 41.32 | Improved |
| use GLO instead of VAE (ablation) | GeLaTO · Eyeglasses few-shot reconstruction (N=3) | PSNR: 35.59 -> 37.19 | Improved |
| stack per-proxy deep buffers before the U-Net instead of z-buffering (DNR) (ablation) | GeLaTO · Eyeglasses view interpolation | PSNR: 41.21 -> 41.32 | No clear effect |
| stack per-proxy deep buffers instead of z-buffering (ablation) | GeLaTO · Eyeglasses few-shot reconstruction (N=3) | PSNR: 36.14 -> 37.19 | Improved |
| stack per-proxy deep buffers instead of z-buffering (ablation) | GeLaTO · Eyeglasses few-shot reconstruction (N=3) | Mask IoU: 0.8725 -> 0.9012 | Improved |
| pretrain a category-level generative model and fine-tune instead of training DNR from scratch (N=30) | GeLaTO · Eyeglasses unseen instance reconstruction | PSNR: 38.75 -> 41.61 | Improved |
| fit all network parameters instead of only latent code z (ablation) | GeLaTO · Eyeglasses few-shot reconstruction (N=3) | PSNR: 31.30 -> 37.19 | Improved |
| fit transformed latent w instead of z (ablation) | GeLaTO · Eyeglasses few-shot reconstruction (N=3) | Mask IoU: 0.7242 -> 0.9152 | Improved |
| fit all parameters instead of only w (ablation) | GeLaTO · Eyeglasses few-shot reconstruction (N=3) | Mask IoU: 0.9152 -> 0.9012 | Hurt |
| also fit texture generator parameters beyond w (ablation) | GeLaTO · Eyeglasses few-shot reconstruction (N=3) | PSNR: 36.50 -> 37.12 | Improved |
| also fine-tune the neural rendering network beyond texture generators (ablation) | GeLaTO · Eyeglasses few-shot reconstruction (N=3) | PSNR: 37.12 -> 37.19 | No clear effect |
| use GeLaTO instead of NeRF with few views | NeRF · Eyeglasses unseen instance reconstruction (N=3) | PSNR: 31.20 -> 36.53 | Improved |
Closest papers by meaning
- Leveraging 2D Data to Learn Textured 3D Mesh Generation (CVPR 2020 · similarity 0.88)
- Single Mesh Diffusion Models with Field Latents for Texture Generation (CVPR 2024 · similarity 0.87)
- TUVF: Learning Generalizable Texture UV Radiance Fields (ICLR 2024 · similarity 0.87)
- Structured 3D Latents for Scalable and Versatile 3D Generation (CVPR 2025 · similarity 0.86)
- MeshGen: Generating PBR Textured Mesh with Render-Enhanced Auto-Encoder and Generative Data Augmentation (CVPR 2025 · similarity 0.86)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
